Raw microbolometer output is unusable: every pixel responds slightly differently and every pixel drifts with temperature. The image you actually see has passed through four stages — NUC, FFC, AGC and detail enhancement — and almost every “why does my thermal look like this” question resolves to one specific stage doing exactly what it was designed to do.
Key takeaways
- NUC and FFC make pixels agree with each other. AGC and DDE decide how the agreed values are displayed.
- Fixed-pattern noise — faint grids or streaks that stay still while the scene moves — means correction is stale or failing, not that the sensor is broken.
- AGC is why a person vanishes when a hot chimney enters frame: the display mapping stretched to accommodate the new extreme.
- Radiometric payloads record temperature upstream of every display stage. Palettes and enhancement change what you see, never what was measured.
On this page
The four stages, and what each one fixes
| Stage | Defect it corrects | When it runs | Visible if it fails |
|---|---|---|---|
| NUC | Per-pixel gain and offset spread from manufacturing | Factory calibration, applied continuously | Coarse mottling or dead-looking pixels |
| FFC | Drift accumulated since power-on | Periodically in flight, shutter closes | Fixed-pattern grids or streaks |
| AGC | 14-bit scene range vs 8-bit display | Every frame | Scene washes out or goes flat when something hot enters |
| DDE | Low local contrast in flat scenes | Every frame | Image looks soft, or noisy if pushed hard |
The order matters. Correction happens before display processing, and radiometric measurement is taken before display processing too — which is why palette and enhancement settings can never affect a temperature reading. Everything downstream of AGC is presentation.
NUC and FFC: making pixels agree
Non-uniformity correction applies factory-calibrated per-pixel gain and offset maps, so that a genuinely uniform scene renders as a uniform image. This is a fixed characterisation of the array, measured against reference sources at manufacture, and it does not change in the field.
What it cannot handle is drift. As the sensor body warms in flight, residual per-pixel offsets accumulate. Flat-field correction closes a shutter for a fraction of a second, presents every pixel with an identical uniform reference, and re-zeroes the offsets against it. The mechanism is covered in detail in the thermal calibration chain.
Skipped or failing FFC shows up as fixed-pattern noise: faint grids, columns or streaks that stay put while the scene moves beneath them. The diagnostic is exactly that — pattern that does not move with the image is sensor-side, pattern that does move is in the scene.
AGC: fitting 14 bits onto an 8-bit screen
The detector senses a far wider temperature span than any display can show. Automatic gain control chooses which part of that span maps onto the visible grey levels, and it re-chooses continuously as the scene changes. This is the stage responsible for the behaviour operators find most confusing.
| AGC mode | What it optimises | Best used for | Failure mode |
|---|---|---|---|
| Full-scene automatic | Contrast across the whole frame | General search and navigation | A single hot object flattens everything else |
| Region of interest | Contrast inside a chosen box | Tracking a target across a varied scene | Ignores what happens outside the box |
| Manual span | A fixed, repeatable mapping | Survey work that must be comparable | Loses detail if the scene leaves the set span |
| Locked after set | Stability during a measurement | Radiometric inspection passes | Must be re-set when conditions change |
The disappearing person. A person tracked clearly at 300 m vanishes the moment a hot chimney enters the frame. Nothing failed — full-scene AGC re-stretched the mapping to include a source hundreds of degrees hotter, and the few degrees separating a person from the background collapsed into one grey level. Region-of-interest or manual span exists precisely for this. It is the first of the seven mistakes operators keep making.
Detail enhancement and palettes
Digital detail enhancement sharpens local edges in low-contrast scenes. It buys visible structure in flat imagery — a uniform roof, a calm water surface — at the cost of amplified noise, and pushed hard it produces halos around genuine edges that can be mistaken for thermal gradients.
The last stage maps values to a palette. White hot, black hot and ironbow present identical data three ways; the choice affects readability and nothing else. Ironbow makes small gradients obvious and makes absolute judgement harder; white hot is the safer default for search. The palette guide covers when each earns its place.
Enhancement and palette are display choices. Neither one touches the radiometric data. On a radiometric payload the per-pixel temperature is recorded upstream of AGC, DDE and palette mapping, so a finding can be re-analysed later with different display settings and the numbers do not move.
Diagnosing an image complaint, in order
- Does the pattern move with the scene? If not, it is NUC or FFC — force an FFC and look again.
- Did brightness change suddenly when something entered frame? That is AGC. Switch to region of interest or lock the span.
- Are edges haloed or the image gritty? Reduce detail enhancement before blaming the detector.
- Is the whole image low-contrast with nothing wrong technically? Check the scene, not the payload — thermal contrast may genuinely be absent at this time of day.
- Only after all four: suspect the hardware.
Running this order costs a minute and resolves the large majority of complaints without a support ticket. It also builds the habit of separating what the sensor measured from how it was displayed — the distinction that underlies radiometric measurement entirely.
Related reading
- Radiometric vs Non-Radiometric Thermal: Which Do You Need?
- Thermal Palettes: White Hot, Black Hot and Ironbow, Compared
- Seven Thermal Imaging Mistakes Drone Operators Keep Making
- Thermal Sensitivity (NETD) Explained for UAV Payload Buyers
- How Uncooled VOx Microbolometers Work
- Shutters, Blackbodies and Two-Point Correction: The Thermal Calibration Chain
- Radiometric Temperature Measurement on UAV Payloads
- MV-2P — 640×512 thermal micro payload
- OP-125A — 30x zoom EO/IR pod with thermometry option
FAQ
Why did my thermal image suddenly change brightness?
Automatic gain control re-mapped the scene when something notably hot or cold entered the frame. The data did not change, only the mapping onto display levels. Lock the span or use region-of-interest AGC when consistent rendering matters.
Does detail enhancement affect temperature readings?
No. Measurement uses radiometric data taken upstream of all display processing. Enhancement, palette and AGC change what you see, never what was measured.
What causes faint grid or streak patterns in the image?
Fixed-pattern noise from stale or failing flat-field correction. The test is whether the pattern moves with the scene: if it stays still while the image pans beneath it, force an FFC.
Questions about the technology? Talk to our engineers — we reply within 2 business days.